As was foretold, we've added advertisements to the forums! If you have questions, or if you encounter any bugs, please visit this thread: https://forums.penny-arcade.com/discussion/240191/forum-advertisement-faq-and-reports-thread/
Options
Supercool at room temperature.
Mad_Morlock
Registered User regular
Registered User regular
I'm linking this thread from Gaming & Technology and drop it here. At this point it seems more appropriate a place to be carrying on a dialogue.
We've established that using a radial peltier to cool a point source of heat will work. There's already a patent that covers that, but mostly applies to LED construction. However, this should prove that moving heat away from point sources will definitely provide a cooling solution. There's is cylindical, mine is cubic. But if we extrapolate the results of their point cooling solution, it goes to say that a second shell of coolers would have similar results. Your hot point has a larger radius, but then so does the periphery of the second outer shell.
We've also established that a peltier shows a better rate of heat transfer if pulsed instead of operating with a steady state current. However, it's previously been established that the operating efficency of a semi-conductor based peltier junction decreases at operating temperature extremes. The optimum operating range for a peltier device is ambient.
So, if we're using the device to only maintain an ambient temperature instead of providing an extreme cooling solution, a regulated pulse current supplied to an array of peltier shells around the processor should provide with an increased surface area to radiate the heat from while maintaining an optimum operating temperature inside the core processor.
The supplied documents:
Thermal Processing Unit - provided by newdor.com media inc.
The radial peltier array patent
The supercooled regulated pulse peltier experiments
Despite the variety of arguments that have been throw against this design, I'm more confident that ever that this will work. Some of the supplied arguments even turned around to suggest that not only is this possible, but that it's already been verified by a number of experiments.
I'd also like to throw this in. But just for kicks.
We've established that using a radial peltier to cool a point source of heat will work. There's already a patent that covers that, but mostly applies to LED construction. However, this should prove that moving heat away from point sources will definitely provide a cooling solution. There's is cylindical, mine is cubic. But if we extrapolate the results of their point cooling solution, it goes to say that a second shell of coolers would have similar results. Your hot point has a larger radius, but then so does the periphery of the second outer shell.
We've also established that a peltier shows a better rate of heat transfer if pulsed instead of operating with a steady state current. However, it's previously been established that the operating efficency of a semi-conductor based peltier junction decreases at operating temperature extremes. The optimum operating range for a peltier device is ambient.
So, if we're using the device to only maintain an ambient temperature instead of providing an extreme cooling solution, a regulated pulse current supplied to an array of peltier shells around the processor should provide with an increased surface area to radiate the heat from while maintaining an optimum operating temperature inside the core processor.
The supplied documents:
Thermal Processing Unit - provided by newdor.com media inc.
The radial peltier array patent
The supercooled regulated pulse peltier experiments
Despite the variety of arguments that have been throw against this design, I'm more confident that ever that this will work. Some of the supplied arguments even turned around to suggest that not only is this possible, but that it's already been verified by a number of experiments.
I'd also like to throw this in. But just for kicks.
Mad_Morlock on
0
Posts
took out her barrettes and her hair spilled out like rootbeer
What a drag.
No dice.
Billiards.
Alternately, is this some sort of free-energy thing? The description means absolutely nothing to me, other than "Peltier".
Really? Wow.
I don't see how it would be site-whoring. It sounds like he came up with an idea and wants to see if it's viable. So let's discuss that. Is it viable? If not, why not?
But it's not site-whoring, it's thread-whoring.
Peltier's also work best at ambient temperature ranges. Very little temperature potential difference between the two junctions, and in an above zero operating range.
If you want to run something way below zero, your best option is the mini-fridge built out of superconductors, insulators and regular conductors.
In the experimental results of the peltier current pulse, it was demonstrated that following a current pulse 6 times the standard steady state operating current of a peltier, it stops working completely for a few moments afterwards. One word for you: depletion.
Finally... if a peltier is only trying to generate a one degree difference between two points, don't you think that requires less energy than generating a full 70 degree difference?
A Peltier is more of an equilibration device. In the presence of no temperature differential, it won't move heat around at all. The bigger the temperature the differential, the faster it moves heat.
The point of using a Peltier in CPU cooling is that it's very good at speeding up thermal transfer. A peltier will get heat away from the CPU very quickly, and generally distribute it over a larger surface area in the process. But, unless you have a sufficient temperature differential between the Peltier and it's environment, then you can actually end up with worse cooling then not having a Peltier at all.
It's like trying to build a mountain out of a processor. All the heat flows downhill through the peltier lattice. But the peltier would have to come in contact with as much of the processor as possible. That's just simple engineering.
A mechanical fan and heatsink cooling combo has already shown to be very effective at maintaining a peltier at equilibrium.
Keep the heatsink, redesign the peltier and processor geometry for maximum coverage and replace the fan with an ionizer.
No moving parts at all. The ionizer works well enough to replace the fan, but the combination of the three should provide for a seamless solution.
Seamless is important. Right now using a peltier on a CPU has major condensation issues.
Workingmen of all countries, unite!
A Peltier does not let you get around this problem, what it lets you do is increase the rate at which heat is removed from one side of the device. but you still have to actually get rid of that heat from the other side, and then some.
Finally, I don't get why you think RT operation will fix anything? The point of cooling in an overclock situation is that you want the CPU to be physically as cold as possible (within limits, throw it in LN and it'll just not work because you've fucked the mobile charge carrier concentrations). The reason you want this is because you need to minimize the amount of thermally activated electron movement. So RT operation, no matter how good, won't help a damn thing in this department.
I've been trying to see if someone can formulate a convincing argument against me.
What the hell? Of course I'm advocating room temperature operation. If I'm going to be in the same room I'd rather it wasn't a freezer. You've got to be able to breather the same atmosphere that's cooling your CPU.
My original point was that if you place the cold side of a pelter in the center of the peltier array, you're translating heat outwards more efficently. That was already demonstrated by the two Russian scientist in the aforementioned patent. Theirs was a radial solution because they're dealing with point sources of heat. Mine is angular because I'm dealing with a heat source with more dimensions.
First off, I'm not trying to overclock. I'm trying to achieve maximum stablility within design specs first.
And overclocking doesn't require the processor be colder as much as you can remove heat from the CPU superlattice before it has the chance to futz the whole mess. Removing heat faster rather than removing it beforehand.
(b) is not a bad idea - the only problem with it that sticks out to me is that it leaves you with an oddly sized 'hot side' to cool, which will make it hard to find an appropriate heatsink. That could be overcome, of course, if this method of cooling was offered as a commercial solution since an appropriately sized heatsink would be included.
On the other hand, (a) is an old and, to be blunt, terrible idea. The problem is that Peltier's don't just transfer heat, they generate more in the process. A peltier with a transfer rating of 100W, for example, transfers (at best) 100W to the hot side from the cold side, but additionally uses at least as much power as it transfers, almost all of which is dissipated from the hot side. The result is that, in transferring N watts of heat energy (N < 100), you end up with at least 2N watts on the hot side. Stack another peltier - with a transfer rating of say 200W, and you end up with at least 4N watts on the hot side of that one, and so on.
It is completely unclear what you're out to achieve. The design specs say "keep it in this range" - if a heatsink and fan can do that then who cares?
Clipse's points about the Peltier are the same ones I've been making, but more directly stated.
And you're idea on what heat does in a CPU is just plane redefining the wheel. Heat = fucking things up. Absence of heat = cold. An overclocked CPU needs to be cold in order to work for various reasons related to how semiconductors function. You must extract heat because it's also dissipating something like 80W of power in a 1cm^2 area. Keeping it cold thus means moving heat away from it at some rate.
Look, if you want a supercool CPU just get a phase change system and be done with it.
Ok... we'll use your math.
You've got a cubic hot body. Emitting 60W.
The each first stage peltier array is designed to move 10W. It's the weakest link in the chain. 6 junctions arrayed on each cube face each translate away the 10W, dumping it at their periphery, into the thermal shell (call it a thermal capacitor) for a net increase to 120W. However, this is still only 20W per cube face. As long as the surface area of the thermal shell more than doubles that of the original hot body, you're in business.
And the original idea was to create a solid state cooling system. Completely electrical. No moving parts. Peltier, heatsink and ionizer.
Transferring N watts of heat across a Peltier will result in an output of at least 2*N watts. The actual power usage depends on a lot of factors, including the height of the Peltier elements, which would necessarily grow in order to make a thicker Peltier plate.
The market beat you to this by quite a while. Zalman has had a (very expensive!) case out for quite a while that uses heatpipes to transfer CPU, GPU, and PSU heat to the case itself, which is designed to maximize surface area in order to dissipate said heat quickly. Not only is this solution solid state, it's passive.
A 10x10x10mm cube with layers of 5mm thick TEC's has a first layer surface area of 2400mm^2. Second layer is 5400mm^2 so it would be capable of handling the heat disapation plus 11% extra. Third layer is 9600mm^2 so it cannot disapate all the heat. Fourth is 15,000mm^2 and will be even worse at disapating heat. You can see where this is going. Also, as Clipse mentioned, thicker layers to try and get an increase in surface area will increase the energy required (heat) to get the same performance as a thinner TEC. As for your example, you would have a hard time finding a TEC rated higher than 10W in a 10x10mm package. As I pointed out earlier, this is a requirement for this to work at all.
Also, a 10x10x10mm cube processor won't be able to disapate heat from its interior fast enough unless you maintain its surface at cryogenic temperatures. Even then, sustainability depends on the rate of heat generation at the core versus the conduction rate through the insulating layers of glass. These temperatures of course has deleterious effects on the functioning of the processor itself.
Your complicated approach offers no benefits over existing technologies.
To head off one other possibility, AMD has patents on integrated on-die TEC coolers.
The surface area of the first cube is actually 600 mm^2 if it's 10x10x10. 100 mm^2 on a side and 6 sides to a cube, right? Maybe I'm missing something...
If your TECs are 5mm on each side, your new cube is actually 20x20x20, for a surface area of 400 mm^2 on a side, for a total surface area of 2400mm^2.
Which is a factor of 4 increase when only a factor of two is required if the heat is being doubled through joule heating of the peltier. This means that the second hotbody cube will now be storing twice as much heat spread throughout six times the surface area.
If the second array of peltiers, which by necessity must be stronger than the first weakest link in the chain, are 1cm in size, the cube will now be 40x40x40mm. 1600mm^2 on a side, for a surface area of 9600mm^2. This is still 4 times as large as the previous surface area.
Maybe I missed something here, but your math just doesn't work for me.
As for the reference to AMDs patent on peltiers intergrated into the chip package and the effectiveness of heatpipes over heatsinks, please refer to fourth page of the ongoing thread entitled "Supercool Computers & Legos" in Gaming & Technology.
morlock you should post this in SE, you'll get more of a response in terms of what you're looking for
Never did get an answer. I figured the AC would have a greater knowledge of CC regs than Games and Technology.
Games and Techology provided a decent forum for establishing arguments and counterarguments.
Debate and Discourse hasn't proved as fruitful.
I'm loath to drag this into another forum on this board myself, however.
Is there any further debate for this thread?
Or have the critics run into a wall?
I've got a physics professor at the local college who wants to put his name on it too.
Which will make it easier to get NRC funding to build a prototype proof-of-concept.
It's a little known fact, but from 1877 to 1878, during his period of work on the carbon telephone transmitter, Thomas Edison used to stroll into coffee shops and conspicuously set up his blueprints on a table in the corner. He would stand over them, and loudly ask himself; "I wonder what would happen if I did so and so", or "maybe if I corrected this." This would continue until a fellow patron inquired as to what he was doing. At that, he would confront them; "Oh, you'd like it if my wonderful invention didn't work, wouldn't you, you critic! Get lost!"
Edison’s embarrassing secret: He never learned how to use a screwdriver.
We've also got a band from Egypt.
Forgive the minimalist webdesign.
I'm not a webdesigner.
Also, we can't theorize that it's empty if you want to have something inside of it.
Yes. People pay for practical things.
But I'm saying that people a generation from now will look back at this time and will think of us as pretty primitive. I'd say even ten years from now the idea of selling a naked processor will be no more.
Right now it's an industry supply and demand issue. If they don't evolve the product to the point where it's sturdy and stable and longlasting, the industry can continue to make a killing off of an ancient architecture.
As for efficency: A heatsink can store and radiate a lot of heat. Tons of it. Its designed to do that. A processor isn't. There will come a point fairly soon where a passive solution just won't cut it anymore. Call it an extension of Moore's Law.
Your high-end computers, yes, will always need cooling, because they're well.. high end. However, not every device in the world is going to need or will be a 4GHz Pentium X, a lot of devices will be 30-200mhz, so on and so forth.
But Moore's Law will hit a wall eventually. It already has on power requirements repeatedly. Heating was an issue with the original Pentium 2. It'll happen again. That's just common sense.
Performance electronics are what we're talking about right now though. Not the processor for your DS or your cell phone. Not even your laptop.
What are the practical purposes in layman's terms? I've spent my whole life around engineers and mechanics and this is the ultimate end to all ends when it comes to design. I would honestly like to know (no, I'm not screwing with you).
From what I have been able to understand, it's primary use would be in either a vacuum or an extremely low gravity environment. Give us a toy and let us play with it.
Here's the assumptions I'm working with:
1. Peltier elements work most efficently when translating heat from a high concentration to a low one. Not just efficently... they can actually generate power.
2. Thermal dissipation with a peltier works most efficently when dumping into an object with a larger radiant surface area.
Now... practical purposes...
Scramjet engines, (supercompression ram jet) operate from Mach 5 upwards to orbital velocity while still in atmosphere. However, going that fast in atmosphere will make your craft melt, especially the point that's breaking through the atmosphere.
There's a partial solution out there that involves using a leading microwave pulse to superheat the atmosphere to reduce friction, but at Mach 25 (low earth orbit velocity) you'll still see MAJOR heating issues.
The ability to translate heat from the hot spots to the main body of the craft might prevent a complete meltdown and actually make these engines usable.
There's one practical purpose.
If you want something a little more home oriented, how about a thermal mesh made with peltier junctions constructed out of quantum dots to increase the flexibility? You could have a whole new generation of thermal clothing that would keep you warm. (Think new electric blanket.)
There's another practical purpose.
But my main interest is in CPUs.
And no... this isn't designed for a vacuum... nor a low gravity environment. An ionizer requires atmosphere. And low gravity means moving parts aren't as much as an issue as it would be in a high gravity/acceleration situation.
It's just an educated guess on the rate of technological growth in the transistor industry. It's not perfect.
The stumbling blocks for Moore's Law are power and heat. Either the chip bleeds too much power, or outputs too much heat. Either way, these are both issues that need to be dealt with for Moore's Law to keep working. So far it's been repeatedly dealt with.
My corollary to Moore's Law is that peltier junction technology will eventually have to proceed apace with transistor development to combat the heating disspation issues created by using a smaller and smaller form factor.
Suppose you eventually end up with a processor the size of a grain of salt. You're just gonna slap a heatsink on that? Doubtful.